Voynich Manuscript folio 13r, Beinecke MS 408, Yale MS 408 · f13r · Yale Beinecke · Public domain
Short version: Currier A and B are two internally consistent statistical "dialects" of Voynichese. The split is real and measurable, different word endings, different character frequencies, even different handwriting. Whether they are languages, topics, or scribal habits is still open. That they differ is one of the few things almost everyone in the field agrees on.

What Currier found

A WWII codebreaker, working by hand, spotted a pattern that has held up for fifty years.

The Observation

Two styles, not one

Currier noticed that some pages were written in a neat, widely spaced hand, and others in a tighter, more slanted one. To his eye, two different people were writing. The statistical behavior of the text tracked that visual split.

The Signatures

Different endings and frequencies

Certain word endings are common in "B" and almost absent in "A." Some symbol groups repeat heavily in "A" and rarely in "B." The two sections are each internally regular, but follow different rules from one another.

His Caution

"Languages," in quotes

Currier was careful: he called them "languages" but stressed they need not be different languages at all. They could be dialects, subject matter, or different conventions. The label stuck; the caution often gets dropped.

Two hands, by the numbers

Currier worked by hand. With modern methods the split can be measured precisely, and reproduced by anyone.

On a character-level conditional measure, the Currier A and B sections separate with an effect size of d = −1.01 (p < 0.001): a large, statistically clear gap, not a subtle one. The difference is also visible in how fast new vocabulary appears within each section. In other words, the fifty-year-old observation survives reproducible, modern testing, and it is one of the strongest internal-structure signals in the entire manuscript.

What it does, and doesn't, mean

A measured difference is not the same as a reading.

It rules out

A single uniform stream

The manuscript is not one homogeneous text. Whatever it is, it was produced under at least two distinct regimes, a fact any successful theory has to explain.

It leaves open

Why they differ

Two scribes, two topics, two stages of an evolving system, or two encodings would all produce a split like this. The data alone does not yet decide between them. Later paleography (Lisa Fagin Davis) finds as many as five distinct scribal hands.

It does not show

Any meaning

Measuring that A and B differ says nothing about what either one says. We characterize the structure. We do not claim to read it.

See our findings → How we measure → Has it been solved? →
Was this page useful?